I did the voltage mod with 4 diodes because I'm lazy

Can I ask where and if did you read the VDDC supposed to be 2,5v and instead @ 3,3v? Cause that would explain quite a lot about this chip lifetime I suppose..

I followed this datasheet http://old.vgamuseum.info/images/stories/doc/ … tions_dec03.pdf
Pages 9,35,45 and the very intresting page 89 where it says that the core can tolerate up to 6v, I can't veliebe it.
I looked for the minimun voltage for the core at default frequencie of 83Mhz and it was stable at 2.0v.

I suppose such number might be the absolute theorical max from a technical point of view more than a real world one. One of the key point of the Rage XL chip I think to remember was that it worked in both 2,5/3,3v voltages for different part of the chip/memory but I suppose not the internal main Vcore instead of the fixed 3,3v of the previous Pro chip, so I still wonder if a fixed 3,3v might still be high. But on the PCB did you measured the VDDC that entered the chip @ 3,3v or maybe is decreased before going to the chip pin?

I am not an expert, so I drew some conclusions without having a deep knowledge, comparing the original version of Ati with the Chinese version I verified that the original version had two mosfets and two voltage regulators and the Chinese version only one, taking into account that the 3.3v of the PCI connection are connected to the drain of the mosfet and to the power supply of the RAM and then the 5v are connected to the flash and then the source receives around 7v, I deduced that the mosfet gate supplies the power to the processor since it is the only thing left. But I may be wrong.

Of course this is already good to know and I am also not expert, it might be interesting to actually measure which is the real VDDC input voltage. I wonder if considering the time passed from the original design that anyway seems it was already more oriented for a 3,3v main PCI usage it might have had sense using a cheaper design leaving out the 5V only old PCI bus but I ask myself if the 3,3v voltage might have not been reduced before the VDDC pin input on the PCB. Because if the chip VDDC theorically could really be powered @ 3,3v doesn't sound great for two decades old ICs making these even cheaper than I thought.

My memory is getting worse faster than I would like and I forgot that I have a Rage XL AGP made by ATI, so today I checked the voltages on the only mosfet it has and they match what I find on the PCI version, my multimeter It reads 2.49v, so I think it is confirmed that the standard voltage is 2.5v and it also seems clear that it tolerates 3.3v without problems.
We would need someone who had a Rage XL PCI made by ATI and has the detail of taking readings in the two mosfets they have.
From my tests and my point of view, these Chinese Rage XL are a better option than for example an S3 trio 3D, mainly because the price of PCI graphics has skyrocketed to a ridiculous level. And then the fact that with the correct driver the Rage XL has been much more compatible than other cheap options and also offers better performance than the S3, of course it is not a TNT but the price is what it is.

I imagine the original two mosfet PCI version of the Rage XL having the same 2,5v VDDC voltage. Considering also the 250nm core manufacturing I suppose the lower power demand maybe for lower power systems was a benefit compared to the Rage Pro chip (beside the better features the XL chip had). I imagine it'd make sense a cheaper not really original design might have preferred to use an higher voltage to save on the design but I also imagine not all the chips might work that easy @ that Vcore who knows. I tried the bios files on my card but it still can't give signal at the VGA output. The card is detected using a primary AGP card but still can't work no matter what I tried. I wonder if existed some msdos tool to test the Rage chip itself. I remember some similar tool for MPEG2 decoder cards that deeply tested the card in msdos for each PCB components, I wonder if something like that exists.

I'm very sorry that the bios files didn't work for you, it's a shame.
I imagine that in the case of the original PCI version of ati, which seems to work perfectly on old equipment, they used a system similar to the one I use, using only the 5v of PCI and a mosfet is for the 2.5v and another for the 3.3v.
Although it may have a system that, in case of detecting the 3.3v in the PCI slot, disables the regulator and the mosfet dedicated to the 3.3v, surely this is easy to do but I do not have the necessary knowledge to implement something like that and that's a shame.
If you have a multimeter, maybe you can check the voltages of the card on the mosfet, my card had, 3.27v on the gate, 3.34v on the drain and 7v on the source of the P3055LD.

Of course I understand these are very cheap cards still I would have liked to test this chip after the Rage Pro AGP (and the quite slow PCI version). It's interesting that the eeprom flash work, the Rage XL chip is detected by both the atiflash/flashrom tools and Win ME too if with another vga booting. Does one of your cards install a Samsung 60ns SDRAM module or like the one above another brand?

I can try resoldering the capacitors to see if those aren't well installed or maybe the clock IC too who knows.

My two cards have Micron technology memory, one rated for 7ns(143mhz) and the oder one for 6ns (166mhz) but this one works at a maximun of 133mhz so maybe is a factory discard.

386sx, can you share the original bios of your ragexl?, I'm curious and I can also try it.

Unfortunately I found during tests that I saved again from the eeprom the new flashed ones overwriting the same file name in the folder so I lost it. 🙁 Anyway it was based still on the usual old id revisions 36Kb too. I think I will buy a new card, while if I find other bios I'll try them too to see if it might work but I begin to suspect they sent me an untested card. What I don't understand is where it could be broken with such few components.. maybe the video chip itself is simply gone that's why I was wondering if the VDDC theorically @ 3,3v might be a possible reason for this. If the ram module was gone I suppose it would still show artifacts or something. But it's still good that the tool and even Win detect the chip and the card.
I will focus maybe on the mosfet and the clock generator, that I might try to change.

I wish you luck fixing it, because even if they are cheap, it's a shame that it didn't work.
I'm also going to buy two more cards soon because I have a dual PIII 1Ghz that only has pci-x and pci, so any good card for it it's expensive, and another Pmmx 200 with a Matrox Mystique 220, the Mystique is famous but not very useful, it lacks Opengl and a good d3d support, so maybe only really useful for 2D.
I'm still praying to get the help of an user capable of modding the bios, I've opened the original bios and the Chinese bios on an hex editor and all of then say the same part number, so there's something modded on the Chinese bios.
I've seen a tool to overclock S3 and CL cards in msdos so everything is possible.
Edit, found a document that I think explain how to program the mach64 cores, mach64GM (RAGE XL), It has a section that explains how to program the PLL registers to set the CPUCLK, MCLK and VCLK frequencies.
http://old.vgamuseum.info/images/stories/doc/ … guide_mar00.pdf

Other question for these Rage XL PCI (4752) cards are that 3DCIF does not work for these, because of no driver support in older drivers that have working 3DCIF support in them, almost all Rage Pro based cards, including Rage XL AGP are working well with 3DCIF, D3D and OpenGL with correct driver. Only Rage Mobility and Rage XL PCI (4752) cards are not working at the moment with 3DCIF, all other Rage3 based cards are good to go with three supported API.

Rage XL AGP 1X,2X (474D ) - 3DCIF works well
Rage XL PCI 66MHz (474E) - 3DCIF not tested
Rage XL PCI 33MHz (4752) - 3DCIF does not work

Re: Proprietary 3D API's
Re: Proprietary 3D API's
https://gona.mactar.hu/ATI3DCIF/

Rage XL AGP (474D) works with modified 4.11.2573/4.11.2582 drivers.

Is it possible to flash these with AGP or PCI 66MHz firmware or change DeviceID code inside these firmware from 4752 to 474D or 474E ?
Maybe it is possible to fix Rage XL PCI (4752) 3DCIF with bios id change or with other Rage XL AGP or PCI 66MHz vga bios.

Gona and Philscomputerslab also should have these Rage XL PCI (4752) cards in hand.

Edit 1
For OverClock:
I know that Powerstrip 2.78 version should work with RageXL memory oc and even all Rage Pro Tweaker versions does not support Rage XL cards overclock option.
https://www.entechtaiwan.com/util/ps2.shtm

Bios:
I should have view AGP versions of Rage XL AGP that perform diferently, maybe i can dumb bioses from there, to compare, if needed AGP ones?

Edit2
Added Rage2 and all Rage Pro Tweaker versions that i have, but i think for XL they should not work for OC, but i havent have any XL PCI cards also, only AGP ones.
In Rage Pro Tweaker there are other options to tweak performance out of Rage3 cards, easyest it was to disable bilinear filtering, when i was testing Quake2 out.

Rage Pro Tweaker version are Public release 1, Version 1.1 and Version 1.2
https://www.upload.ee/files/14115369/rage2-ra … weaker.zip.html

At the moment with the information I have about the Chinese Rage XL PCI....
Well, there are several details that differentiate it and one of them is the BIOS, from what little I understand it is quite different from the original ATI PCI.
I haven't gotten 3DCIF to work either, the only thing I've tried is the Wipeout version and the only thing it says is that there is no 3DCIF in the system.
From the documentation I've found all the mach64 chips look a lot alike I don't know if changing the id in the bios would be perfect, but it shouldn't be impossible and the chinese PCI version and ati agp are quite similar in some respects in my opinion.
Great research work regarding 3DCIF drivers.
I'll try to do some research myself.
There is a lot of information in that thread about it, thank you very much.

It might be that the difference in the bios compared to the original PCI version might depend also from the combination of video chip revision/silicon/eeprom version/sdram module parameters and where they were installed (just a theory) probably originally in some oem mainboards, who knows which and where and from what I understand to make it functional in an add-in card configuration the bios had to be configured specifically. Maybe if they started from the original mainboard oriented bios the resulting one I imagine might be a sort of "modded" version (I suppose there were different PCI boards too) compared to the few original PCI Rage XLs but probably not like a real optimized PCI oriented version.

It'd be interesting to have some of those old PCI cards and bios to test. Also something I wish would exist should be some low level msdos testing tool I suppose it might have existed to check the chip status. Do anyone know about something like that maybe for the Rage Pro chip too that might work here too. As said lately I tested a MPEG2 decoder card that had its oem official msdos floppy disk tool to test the low level functions of the chip, the bios, the ram, even testing the decoding capability of the card. I suppose existing cause the config combination of where the card could work correctly weren't really many.

Ati Rage3 cards are also Mach64 cards, maybe these can be useful.

64UTLxxx.ZIP/EXE (DOS) - ATI mach64 vx.xx Utilities Installation & Configuration - contains a video memory test, among other things, for ATI mach64.
http://web.archive.org/web/19970404103745/htt … rs/64utl222.zip
http://web.archive.org/web/19970404103755/htt … rs/64utl222.txt
http://web.archive.org/web/19970706062651/htt … rs/64utl303.exe
http://web.archive.org/web/19970706062702/htt … rs/64utl303.txt

Info taken from here:
Re: Testing VRAM on older cards under DOS?

Thanks, I'll try these. Meanwhile I was reading about many Dell PowerEdge computer models having Rage XL cards with the same/similar bios versions. Not sure if they were integrated or dedicated cards all the times but I suppose these ICs originally came from these systems or even some mini PCI IBM add in modules that had the chip and the sdram module (but strangely not the bios eeprom from what I see online).